Biodiversity Trends Show an Excess of Both Near Stasis and of Very Large Change

Temporal trends in biodiversity metrics such as species richness and total community abundance often average close to zero across many assemblages. However, little attention has been given to characterising the variation in trends such as the overall shape of the distribution. For a variety of biodiversity metrics and large data sets, we find that measured rates of change are not well fit by the Normal Distribution but are better characterised by leptokurtic, fat-tailed distributions. The best fit overall is to the Subbotin, with the special case of the Laplace Distribution usually performing well. These findings can improve statistical analysis and simulations, but the most important implication of leptokurtosis is that ecologists should pay more attention to the small number of systems that are experiencing strong temporal trends, rather than focusing on the large majority of cases that exhibit weak trends.